Third Wave Software - Part 1: Background
This is the first of two posts that attempt to describe salient aspects of software in the third wave of computing. An appropriate subtitle might be: "Microsoft is out of its league invading the embedded systems space." Even more astonishing, Microsoft's inability to deeply penetrate the embedded systems space will be its undoing, much like IBM was undone by Wintel. Microsoft will continue to hold sway over PC's and most servers, but no longer will it prevail as the most influential force in computers. The Microsoft hegemony will be undone by billions of hidden computers that follow a different drummer.
The key to understanding tidal forces causing primary waves in the computer industry is to look objectively at sources of control in prior waves, and to extend trends already evident in the beginnings of the next wave, in this case hidden computing. Recall the first two primary waves of computer processing, represented respectively by the mainframe and PC paradigms. The mainframe/minicomputer wave peaked at the end of the seventies, while the PC paradigm currently is in full flower. The biggest wave yet, the third wave has barely begun, and consists of hidden computers, the realm of embedded systems. This is all old hat on this thread, but what we haven't discussed before, and what should be understood, is how the roles of hardware and software evolved in parallel with the first two waves, and how they can be expected to evolve in the third wave.
Mainframe and minicomputer companies produced computers that vertically integrated proprietary hardware, peripherals, operating system and software tools. Centralized IT departments took responsibility for application programs. This proprietary tying together of hardware and software continued well into the eighties, despite the fact that the antitrust police forced IBM software to be unbundled from hardware. Third party enabling software was not a significant factor during the first wave. In fact, software and customer support were necessary inconveniences enabling the selling of "Big Iron".
Probably thanks to concerns about antitrust issues, IBM divvied up the PC operating system to Microsoft and the microprocessor to Intel. IBM's uncharacteristic decision to open the PC paradigm to third party developers resulted in an exploding industry of add-on hardware and software products and clones. The dominance of Wintel over Apple came about because third party developers were encouraged to develop add-on products around the novel, open notion of shrink-wrapped software and compatible hardware add-ons, and Microsoft had the good sense to limit MSDOS, and early incarnations of Windows, to the Intel Architecture. The important point for us to note is that the emergence of the second wave occurred along with the separation of hardware and software.
While IBM was the catalyst creating Wintel, everything about the model, the openness, the separation of hardware and software, everything, was anathema to IBM. IBM tried to choke off this insanity and take back control with the introduction of a proprietary Micro-Channel Architecture (MCA) Bus and OS/2. The budding clone industry rebuffed IBM on the MCA Bus, and Microsoft eventually quit the OS/2 effort, taking with them their Windows marbles and market penetration. Both actions forever ended IBM's parental role with regard to the PC paradigm. More important, these events left the world with a legacy of open systems, and with hardware segmented from software.
You might argue the separation of hardware and software in the Wintel consortium was due to happenstance and not at all material to the ultimate dominance of the IBM-Compatible PC (notice it is not called IBM compatible any longer). Openness to third party innovation lent more to the ultimate dominance of the IBM PC than separation of hardware and software. While that may have been true early in the game, the separation of hardware and software became critical later, facilitating the spread of Windows NT and CE on non-Intel platforms, as well as the spread of Intel processors based on operating systems other than Microsoft's.
Once the Wintel PC monopoly was in place, Intel and Microsoft both expanded by designing out competition. Intel integrated various hardware functions into processor chip sets, and extended its domain from processors to the motherboard, with little fear of retaliation from computer manufacturers. To this day, Microsoft inexorably expands the PC operating system to include capabilities provided by existing software vendors. With monopolistic pricing assured for the "Intel Inside" processor and for the "Windows" operating system, each company has economies of scale and leeway to price add-ons cheap enough to defeat competition, as illustrated by a long list of companies that suffered in competition with these companies.
As we turn to the emerging embedded systems space, and focus first on hardware, we see that Intel has had no success whatsoever in extending its hegemony to the chips inside smart devices. Intel has a formal initiative to capture the major share of high-end, server computers (the first wave remains a sizeable business), but seems willing to be a mere player in embedded processors. Why? The reason is simple. The microprocessor and its kin (DSPs, microcontrollers, ASIC, Integrated Circuits housing Systems-on-a-Chip) are commodities, at least at the "processor" level. Intel can compete in a commodity business, but it cannot exercise monopoly power, the basis of its huge market cap. (To appreciate both Intel's and Microsoft's dependence on monopoly power, don't look at P/E ratios. Look instead at operating margins and net profit rates.)
This reluctance of Intel to aggressively attack the embedded processor space because of commodity pricing begs the issue of whether this is the natural state of embedded processors, or whether Intel just hasn't gotten around to changing things. The sad truth for Intel is the embedded systems space is too varied, and subject to far too many constraints to be dominated by a single processor architecture or semiconductor company, even one the size of Intel. The reason for this is the significant contribution of hardware to total cost, footprint and functionality of particularly high-volume embedded products, and its not going away anytime soon. Lacking the equivalent of Windows to force compliance to embedded Intel architecture, Intel is forced to view embedded processors as a commodity market; albeit a market in which they must participate. I suspect Intel was distraught when Microsoft introduced Windows CE with the accompaniment of a number of non-Intel platforms. Any chance of extending the Wintel juggernaut into the embedded space ended abruptly that day.
The fact that embedded processors are, and will remain, commodities does not mean hardware platforms will not contain Intellectual Property (IP). There is a whole new meaning emerging for IP in hardware, but it is occurring at a lower level than the processor as a whole. Its called Semiconductor IP (SIP), and it consists of so-called hard or soft cores needed to define modules used in complex chips. Companies like Rambus own the IP needed to do something special on a chip. Advanced RISC Machines (ARM) provides a CPU core used by a number of semiconductor companies, two of which have announced I2O chips based on ARM processors. Even IBM will help customize a system-on-a-chip with modules of your choice coupled with a PowerPC CPU core. SIP companies license the module, conceivably at various design levels (hard, soft, firm, etc.), to semiconductor companies. Should the proprietary core become a de facto standard, the company stands to reap huge rewards through royalties.
Now think about it. First wave monopoly power IP consisted of the total package of all hardware and enabling software, operating system and utilities. By the second wave, IP bifurcated into hardware and software, but each formed around a smaller "tollgate" portion of the applicable domain: the microprocessor for hardware and the entire OS for software. Simplicity needed to fuel a mass market dictated near universal subordination to these symbiotic tollgates. (Microsoft extending its monopoly in office application software by using economies of scale and "tying", the latter probably legal but upsetting to the DOJ and others.)
It is evident that hardware IP in third wave will evolve at the sub-processor level, to so-called cores generally provided by SIP companies like Rambus, not semiconductor companies themselves. Further, the variety of hardware components should continue to expand rapidly, adding complexity to hardware design but justified by ever-decreasing unit costs of product production. Lower size and costs coupled with increased capabilities is a formula for exploding functional requirements, a formula that already is forcing software to metamorphose to simultaneously thwart complexity while somehow maintaining adequacy for an expanding list of tasks.
At this point you should be sensing that something about Microsoft's tidy world doesn't fit well in embedded systems, although you can't quite decide what, or exactly what it means. You have probably already noted that both Intel and Microsoft are dedicating much more time and effort to swimming up the server stream than floating down the PC river to the ocean of embedded systems. You now know why this makes sense for Intel, but what about Microsoft?
The next post focuses on the changing face of software in the third wave.
Allen |
